Attention Deficit Hyperactivity Disorder (ADHD) is a chronic and impairing condition affecting as many as 3-5% of children. It is now widely suspected that the diagnostic category of ADHD encompasses children with a range of etiologies and possibly with distinct pathophysiologies. Thus, the field has sought improved intermediate phenotypes that can be used as targets for assessment, gene finding, mechanism, and to improve characterization of heterogeneity. Reaction time (RT) speed and variability are considered highly promising as low-cost cognitive endophenotypes for ADHD, and also show promise as trans-diagnostic phenotypes that could be applied to a range of psychiatric disorders. However, modest effect sizes remain a fundamental obstacle to the power of even these leading endophenotypes. These modest effect sizes result from at least two problems. First, traditional analyses of RT variables fail to accurately measure the specific mechanisms that drive group differences. Effect sizes remain modest because they continue to reflect the effects of multiple processes, only some of which differ between groups. Second, it is unclear whether one of these mechanisms accounts for slow, variable RTs for all children with ADHD or whether different sub-groups of children have distinct profiles. If subgroups exist, power to use RT variables as an endophenotype will be muted until these groups are identified. These problems are not tractable with the models of RTs typically applied because they do not disentangle the multiple processes impacting the final RT. The current project seeks to solve this problem and to isolate the reason for slow, variable responding in ADHD, to enable better conclusions about its endophenotype promise and, ultimately, to better characterize heterogeneity within the disorder. To do so the applicant will apply a neurally-informed model of speeded decision-making, called a drift diffusion model. Using two large, independent datasets, she will evaluate which mechanisms account for group-level differences in RT variables between ADHD and non-ADHD youth (Aim 1). The applicant will also evaluate the familiality of the effects by comparing parents and siblings of children with ADHD to those of non- ADHD controls (Aim 2). Finally, she will use latent transition analyses applied to a longitudinal datase, to determine whether one mechanism accounts for slow, variable RTs for all children with ADHD or different sub- groups exist (Aim 3). The stability of subgroups across time will also be assessed and compared to the stability of diagnostically-defined subgroups. There is already evidence that cognitive heterogeneity is related to individual differences in impairment for those with ADHD. Successful identification of a more homogenous and interpretable cognitive endophenotype for ADHD will provide a new, easily disseminated target for genetic studies and for potential re-assessment of heterogeneity. Results will also inform future efforts at refining nosology using trans-diagnostic phenotypes, thus addressing aims of the NIH Research Domain Criteria initiative.